Merge patch series "scsi: target: Allow userspace to config cmd submission"

Mike Christie <michael.christie@oracle.com> says:

The following patches were made over Linus's tree but apply over
Martin's branches. They allow userspace to configure how fabric
drivers submit cmds to backend drivers.

Right now loop and vhost use a worker thread, and the other drivers
submit from the contexts they receive/process the cmd from. For
multiple LUN cases where the target can queue more cmds than the
backend can handle then deferring to a worker thread is safest because
the backend driver can block when doing things like waiting for a free
request/tag. Deferring also helps when the target has to handle
transport level requests from the recv context.

For cases where the backend devices can queue everything the target
sends, then there is no need to defer to a workqueue and you can see a
perf boost of up to 26% for small IO workloads. For a nvme device and
vhost-scsi I can see with 4K IOs:

fio jobs        1       2       4       8       10
--------------------------------------------------
workqueue
submit        94K     190K    394K    770K    890K

direct
submit       128K     252K    488K    950K    -

Link: https://lore.kernel.org/r/1b1f7a5c-0988-45f9-b103-dfed2c0405b1@oracle.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

2 files changed, 22 insertions, 7 deletions

diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index 159567359bbb..97099a5e3f6c 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -108,6 +108,15 @@
 #define SE_MODE_PAGE_BUF			512
 #define SE_SENSE_BUF				96
 
+enum target_submit_type {
+	/* Use the fabric driver's default submission type */
+	TARGET_FABRIC_DEFAULT_SUBMIT,
+	/* Submit from the calling context */
+	TARGET_DIRECT_SUBMIT,
+	/* Defer submission to the LIO workqueue */
+	TARGET_QUEUE_SUBMIT,
+};
+
 /* struct se_hba->hba_flags */
 enum hba_flags_table {
 	HBA_FLAGS_INTERNAL_USE	= 0x01,
@@ -717,6 +726,7 @@ struct se_dev_attrib {
 	u32		unmap_granularity;
 	u32		unmap_granularity_alignment;
 	u32		max_write_same_len;
+	u8		submit_type;
 	struct se_device *da_dev;
 	struct config_group da_group;
 };
diff --git a/include/target/target_core_fabric.h b/include/target/target_core_fabric.h
index b188b1e90e1e..3378ff9ee271 100644
--- a/include/target/target_core_fabric.h
+++ b/include/target/target_core_fabric.h
@@ -113,11 +113,20 @@ struct target_core_fabric_ops {
 	struct configfs_attribute **tfc_tpg_nacl_param_attrs;
 
 	/*
-	 * Set this member variable to true if the SCSI transport protocol
+	 * Set this member variable if the SCSI transport protocol
 	 * (e.g. iSCSI) requires that the Data-Out buffer is transferred in
 	 * its entirety before a command is aborted.
 	 */
-	bool write_pending_must_be_called;
+	unsigned int write_pending_must_be_called:1;
+	/*
+	 * Set this if the driver supports submitting commands to the backend
+	 * from target_submit/target_submit_cmd.
+	 */
+	unsigned int direct_submit_supp:1;
+	/*
+	 * Set this to a target_submit_type value.
+	 */
+	u8 default_submit_type;
 };
 
 int target_register_template(const struct target_core_fabric_ops *fo);
@@ -166,20 +175,18 @@ int	target_submit_prep(struct se_cmd *se_cmd, unsigned char *cdb,
 		struct scatterlist *sgl, u32 sgl_count,
 		struct scatterlist *sgl_bidi, u32 sgl_bidi_count,
 		struct scatterlist *sgl_prot, u32 sgl_prot_count, gfp_t gfp);
-void	target_submit(struct se_cmd *se_cmd);
+int	target_submit(struct se_cmd *se_cmd);
 sense_reason_t transport_lookup_cmd_lun(struct se_cmd *);
 sense_reason_t target_cmd_init_cdb(struct se_cmd *se_cmd, unsigned char *cdb,
 				   gfp_t gfp);
 sense_reason_t target_cmd_parse_cdb(struct se_cmd *);
 void	target_submit_cmd(struct se_cmd *, struct se_session *, unsigned char *,
 		unsigned char *, u64, u32, int, int, int);
-void	target_queue_submission(struct se_cmd *se_cmd);
 
 int	target_submit_tmr(struct se_cmd *se_cmd, struct se_session *se_sess,
 		unsigned char *sense, u64 unpacked_lun,
 		void *fabric_tmr_ptr, unsigned char tm_type,
 		gfp_t, u64, int);
-int	transport_handle_cdb_direct(struct se_cmd *);
 sense_reason_t	transport_generic_new_cmd(struct se_cmd *);
 
 void	target_put_cmd_and_wait(struct se_cmd *cmd);
@@ -197,8 +204,6 @@ void	target_stop_session(struct se_session *se_sess);
 void	target_wait_for_sess_cmds(struct se_session *);
 void	target_show_cmd(const char *pfx, struct se_cmd *cmd);
 
-int	core_alua_check_nonop_delay(struct se_cmd *);
-
 int	core_tmr_alloc_req(struct se_cmd *, void *, u8, gfp_t);
 void	core_tmr_release_req(struct se_tmr_req *);
 int	transport_generic_handle_tmr(struct se_cmd *);